For Developers, feel voice providers like VAPI wouldn't be required in near future. Directly integrate the OpenAI API, and have components like WebRTC, real time streaming, client server connection mapping, DB connections & data mapping implemented. For handling workflow management, state management, could integrate certain frameworks on top like Langraph.
Those platforms are already not required anymore, but I believe the realtime API will be the reason they become even more popular. Will share more on that soon.
Great video as always! Since you are probably in contact with the Vapi team... Can you estimate how long it will take until this is implemented? Thanks.
Great video. I wonder what the future will be like with Voice AI becoming this realistic. How long do you think it will take for Vapi to implement this? (few days, weeks, months?)
I haven’t even started with voice AI education. Honestly, for now I’m happy helping others build out extremely powerful systems, but the educational route might certainly be interesting one I see the need for it
What a great video! Thanks so much for doing the work and providing us with the template for free. If you don’t mind me asking, how can I reduce my costs with Twilio and set up an open-source phone system to act as the call gateway? Another thing I was planning to implement WebRTC as it has the functionality to reduce Eco and noise reduction in case someone will call in a loud environment!
I think OpenAI handles the noise reduction part by themselves. If you're referring to SIP trunking, you most likely need to see how you can do the connection. Not every platform allows you to add a SIP URL to it, sometimes it's the other way around. If you want to try it, use something like Zoiper
You are as always authentic in your opinion. It is exciting thing for someone who is beng introduced to this voice stuff with ai for the first time. What do you thinkis the basic thing a beginner can learn in low code development ? What is the skill that moves the needle?
Understanding the concept and foundations. I think that’s the most important thing. Try some of my examples so you have a working solution, and then try to understand how it’s done. That’s a great point to start. 👍🏻
İt's just way too expensive. Some people payed $3 for 5 minutes, even though the pricing catalogue says it's around 30 cents a minute... Simply unacceptable
You can achieve the same with Vapi by dropping 50k tokens into your master prompt :) Anyways, API costs will definitely come down, so that isn’t a concern in my opinion
SO could I build a conversational chat app. Basically give someone a person to talk to as they walk around and chat with? are prices too limiting right now??
We haven't tried that yet, but if you have a number, you can most likely make calls to it through a provider like Twilio. There are also other approaches that you might be able to leverage long term, such as daily.co
Are you sure you watch your videos on normal playback speed? :D I've mentioned some of the benefits in the video. If that's not enough, I'll drop a more detailed one soon.
I can see what causes your disappointment. You'll always see major languages being implemented at a faster pace. Honestly, I'm already impressed it properly handles multilingual conversations as smooth as now, as this was already incredibly hard with the orchestration layers we've seen so far. We should be happy about those advancements and help them with enough input to make it even better, which on the other hand will also increase your chances of having better results for other languages.
@@jannismoore Indeed, I am disappointed, as I have experience applying language models in IVR systems since the 2000s, and I understand that implementing a new model in 2024 should not pose a problem. The underlying issue seems to be a lack of concern for anything outside a specific "cultural" circle. In summary, they simply don't care. But I do hope I am mistaken. I truly appreciate your videos; they bring a refreshing perspective to this emerging area .
Great video, thank you, Jannis!
Great video as always Jannis, let's go 🔥
Very informative, will start jumping in, thanks for the free resources.
Was waiting for you to release this!
Love its speed, unmatched by anything else out there right now!
Thanks for the helpful video Jannis!
For Developers, feel voice providers like VAPI wouldn't be required in near future. Directly integrate the OpenAI API, and have components like WebRTC, real time streaming, client server connection mapping, DB connections & data mapping implemented. For handling workflow management, state management, could integrate certain frameworks on top like Langraph.
Those platforms are already not required anymore, but I believe the realtime API will be the reason they become even more popular. Will share more on that soon.
Great video as always! Since you are probably in contact with the Vapi team... Can you estimate how long it will take until this is implemented? Thanks.
I’m not quite sure, but I assume we should see something being released soon.
Great video. I wonder what the future will be like with Voice AI becoming this realistic. How long do you think it will take for Vapi to implement this? (few days, weeks, months?)
I expect weeks max. :)
What are your plans Jannis? Run the agency long term, or switch completely to saas, or voice ai education, or something else?
I haven’t even started with voice AI education.
Honestly, for now I’m happy helping others build out extremely powerful systems, but the educational route might certainly be interesting one I see the need for it
What a great video! Thanks so much for doing the work and providing us with the template for free. If you don’t mind me asking, how can I reduce my costs with Twilio and set up an open-source phone system to act as the call gateway? Another thing I was planning to implement WebRTC as it has the functionality to reduce Eco and noise reduction in case someone will call in a loud environment!
I think OpenAI handles the noise reduction part by themselves. If you're referring to SIP trunking, you most likely need to see how you can do the connection. Not every platform allows you to add a SIP URL to it, sometimes it's the other way around.
If you want to try it, use something like Zoiper
You are as always authentic in your opinion. It is exciting thing for someone who is beng introduced to this voice stuff with ai for the first time. What do you thinkis the basic thing a beginner can learn in low code development ? What is the skill that moves the needle?
Understanding the concept and foundations.
I think that’s the most important thing.
Try some of my examples so you have a working solution, and then try to understand how it’s done.
That’s a great point to start. 👍🏻
@@jannismoore that is good to hear.
Can you share your replit link? Thanks
I did! It’s in my resource hub which you’ll find in the description
I'm surprised that vapi isn't on top of this already.
They are :)
İt's just way too expensive. Some people payed $3 for 5 minutes, even though the pricing catalogue says it's around 30 cents a minute... Simply unacceptable
Cost will go down soon just like other API costs
You can achieve the same with Vapi by dropping 50k tokens into your master prompt :)
Anyways, API costs will definitely come down, so that isn’t a concern in my opinion
@@jannismooreexactly. Just a way to raise the bar of the value of their product because they cannot already scale.
SO could I build a conversational chat app. Basically give someone a person to talk to as they walk around and chat with? are prices too limiting right now??
You can do that, but yes, prices are still limiting as of now.
I do believe that those will come down quite rapidly.
Hey there. Thank you for tge value bombs you are dropping...Heads up the link to the resource seem to be broken...thanks again
Appreciate it! Both of the links work when opening them. What do you see once you click on them?
How do we get acces to the code you made? 👍
Via my resource hub - the links for that are in the description :)
could it work with Microsoft Teams Phone? I would like to use it in an IVR setup
We haven't tried that yet, but if you have a number, you can most likely make calls to it through a provider like Twilio. There are also other approaches that you might be able to leverage long term, such as daily.co
Can anyone suggest how this could be used for real-time translation? Actually it doesn't need to be voice to voice just voice to text
In that case you might just want to look at Deepgram
This seems as slow as vapi. What advantages does this, will this have, if it’s the same speed without and of the features of vapi?
Are you sure you watch your videos on normal playback speed? :D
I've mentioned some of the benefits in the video. If that's not enough, I'll drop a more detailed one soon.
Bro this is not slow at all… You do realise it should sound human and not respond in 0.005 milliseconds? The delay makes it sound human smh
Jannis is the ONLY way to go when it comes to AI Voice and Automations.
How can we use it for ai call bots?
You can use the custom example I showed for Twilio, or you can give it another couple of days and Vapi will most likely have something available too
can't wait til this kills customer service phone jobs. calling my wireless carrier for something is often a big trouble, taking hours!
The cost is way too high, we need open source models
I don't think the price will be that high for long
Laughable. It works in English or German, with simple Indo-European languages. But it dies with Hungarian. instantly.
I can see what causes your disappointment.
You'll always see major languages being implemented at a faster pace. Honestly, I'm already impressed it properly handles multilingual conversations as smooth as now, as this was already incredibly hard with the orchestration layers we've seen so far.
We should be happy about those advancements and help them with enough input to make it even better, which on the other hand will also increase your chances of having better results for other languages.
@@jannismooreI hope Dutch works already, I really need a Dutch agent. VAPI starts hallucinating on Dutch and speaking half German after a while lol
@@jannismoore Indeed, I am disappointed, as I have experience applying language models in IVR systems since the 2000s, and I understand that implementing a new model in 2024 should not pose a problem. The underlying issue seems to be a lack of concern for anything outside a specific "cultural" circle. In summary, they simply don't care.
But I do hope I am mistaken.
I truly appreciate your videos; they bring a refreshing perspective to this emerging area .
You need to change your thumbnail. It looks evil
Seems like you clicked on it nevertheless